Not so fast: high-frequency financial data for macroeconomic event studies

Over the last decade, it has become increasingly popular to use event studies with intraday asset pricing data to study the effect of macroeconomic events on the economy. The proponents of this approach argue that asset prices react to macroeconomic events very quickly and that if we know the precise timing of a macroeconomic announcement, a very narrow event window around such an announcement (ranging from 30 minutes to 60 minutes) should be sufficiently long and free from contaminating information that might otherwise cause biased estimates in wider event windows. In contrast, this paper argues that even narrow event windows can lack clean identification because the reaction of asset prices may be affected by other important news that comes out earlier on the same day. We support this argument by studying the relationship between federal funds futures and other asset prices (stocks and Treasuries) on FOMC announcement dates, a relationship widely studied in high-frequency event studies to identify the effect of conventional monetary policy shocks on asset prices. We find that asset prices react significantly more strongly to monetary policy shocks on FOMC announcement dates that overlap with other macroeconomic announcements that come out earlier on the same day. We also find a stronger reaction of asset prices when markets are more volatile. This finding suggests that limitations of investors, such as through rational inattention or asymmetric information, might matter in these event studies. Consequently, one should be cautious before arguing that high-frequency (intraday) event studies adequately address the contamination issues that plague the methods that use low-frequency data.